Function Modules
- Encryption tool: The inference service is provided through the open RESTful API. Two-way certificate authentication is enabled by default. The certificate is provided by the third-party platform integrated with the inference service. The certificate private key supports the certificate import, but its password must be encrypted using the encryption tool provided by the inference service.
- Inference configuration file generation module: Dynamically generates the pipeline files required for creating streams by MindX SDK and the cfg and label files required for postprocessing based on the parameter files and models output by the training component.
- Inference module: Uses MindX SDK to load the configuration file and creates streams of the corresponding service to provide inference capabilities.
- Post-processing module: Post-processes the inference results based on the service scenarios and corresponding parameter configuration. The scenarios include component error, missing, or reverse, industrial OCR, glue detection, foreign object detection, and label detection.
- Module for reporting the inference result to a third-party platform: uploads the inference result to a specified address or reports the inference latency to Prometheus.
- Customized scenario module: Users can expand scenarios based on actual requirements, including customized pipeline orchestration and customized post-processing.